Can Innovation Be Automated?

JimmyQS writes "The Harvard Business Review blog has an invited piece about Innovation Software. Tony McCaffrey at the University of Massachusetts Amherst talks about several pieces of software designed to help engineers augment their innovation process and make them more creative, including one his group has developed called Analogy Finder. The software searches patent databases using natural language processing technology to find analogous solutions in other domains. According to Dr. McCaffrey 'nearly 90% of new solutions are really just adaptations from solutions that already exist — and they're often taken from fields outside the problem solver's expertise.'"

Comments

Following Betteridge's law of headlines

[grouchomarxist]

No.

http://en.wikipedia.org/wiki/Betteridge's_law_of_headlines

Can we stop this tired cliche?

[six025]

No.

http://en.wikipedia.org/wiki/Betteridge's_law_of_headlines

;-)

It's really well past its use by date.

Re:Can we stop this tired cliche?

[TeknoHog]

But correlation is not causation! Also, hot grits.

Re:Can we stop this tired cliche?

[stud9920]

No. If only because of the form of your comment subject

"One Click"

[Required Snark]

Yep, just as creative as Bezos's "One Click" patent. Perfect technology for a legal regime dominated by lawyers with patent examiners recruited from regions that have only horses, and people go to town to use the only telephone.

Just imagine how great it will be when Google, Microsoft, Apple, Oracle, HP, IBM, etc get in a automated patent race where they each file millions of patents applications a month.

They'll just do to patents what they did to taxes; change the rules so that the more you file, the less you pay, and the big players make the government pick up the tab.

Why should intellectual property be any less corrupt then Wall Street? After all, big bank profits are derived from direct subsidies, so why should big tech have to pay for patents? They deserve to be on the corporate gravy train just as much as Goldman and JPMorgan.

Anything else would be unamerican. Don't you want to win the war on drugs, terrorism, the environment, free speech, privacy, ...?

Re:Except the Answer is unfortunately Yes

[gl4ss]

the headline isn't "can research be made more efficient by using machine searches?".

Re:Except the Answer is unfortunately Yes

[rioki]

Except that "Can Innovation Be Automated?" is clearly answered with a bright and clear NO by the article and summary. They did not build a "innovation" software but a "redundant invention finder" software. Except that the word innovation is tossed around willy nilly and is basically a synonym for development (as in the D part in R&D). Note though that it is never used for research, that could actually lead to innovation. Sure the software may be useful, but I think more to a patent attorney, that an engineer. In my 10 year career as software developer I have never seen real innovation, just solving an old problem in a new context and it helps to know it was solved by others, but in the end you need to take the local specialties into account.

TRIZ

[eulernet]

I'm surprised that the article doesn't mention TRIZ and ASIT, which are methodologies for innovation.

TRIZ was invented by Genrich Altshuller in 1946, and has been used by russian engineers to counter the american domination on technology.
http://en.wikipedia.org/wiki/TRIZ

The history behind TRIZ is interesting, since Genrich Altshuller http://en.wikipedia.org/wiki/Genrich_Altshuller was working as a clerk in a patent office (like Einstein), and he noticed that the patents were using some patterns.
He started to categorize all patents to enumerate the used patterns, and he found 39 characteristics with 40 generic solutions.
The idea is that you want to solve a contradiction between 2 characteristics, the contradiction is called a "conflict".
A contradiction matrix of 39*40 cells has been built: http://www.triz40.com/
Recently, the TRIZ group succeeded to verify that the matrix was able to map more than 3,000,000 patents.

TRIZ was kept as a secret before the Soviet Union exploded, then the russian engineers went to a lot of different countries.
In Israel, the TRIZ group started to simplify the methodology in a smaller set, called SIT.
Very recently, Roni Horowitz simplified SIT into ASIT, which is a set of 6 rules able to map innovation.

TRIZ explains that there are 5 levels of invention:
http://www.trizexperts.net/5levels.htm
and it's dedicated to the 4 first levels.
TRIZ is also more adapted to engineers that need a framework to solve problems, but it's not really creative in my opinion.

Re:Except the Answer is unfortunately Yes

[dkf]

In my 10 year career as software developer I have never seen real innovation

Keep on looking! It does happen very occasionally, and it is wonderful to observe. Hold to the hope!

Re:Except the Answer is unfortunately Yes

[dbIII]

Doesn't matter, the requirement these days is that you pay the fee. Everything else is extra. Prior art is seen as a court's problem.

90% of new solutions ...

[Taco Cowboy]

90% of new solution may be, as TFA stated, re-adaption of existing solutions into other fields

But that's not "innovation" in pure sense

Innovation is something that is new

It may be a combination of two old items, like putting tea leafs in a bag made of paper, the result, however, is a brand new thing

That "90%" quote from TFA is akin to replacing "tea" with "coffee" with the outcome of "coffee bag" instead of "tea bag"

Thus, having a software that "innovates" may offer us some "re-application of technologies", but it won't give us new ideas

Re:90% of new solutions ...

[Stirling Newberry]

An idea is joining two things which seem different, but can be shown to actually be the same. By providing possible reapplication, that is part of the way to "something new," depending on how far afield. If you look at many "new" ideas, the parts and their origins become obvious. The "new" part will often be the means of moving the solution to its new context.

Take, for example the derivation of the Lorentz contraction from a description of the movement of light in aether. Lorentz simplified the mathematics by inventing the idea of local time, to move equations meant for kinematics to this new context of Maxwellian radiation. Poincare recognized that "local time" was an ingenious idea, but did not quite get to what we think of as relativity. The Lorentz contraction, and "local time" are then moved, essentially wholesale, into Einstein's kinematics.

New isn't always the elephant, it is the ability to visualize the elephant where it has never been before. Since innovation is not a completely black box problem, aiding visualization of it can be valuable.

Re:90% of new solutions ...

[Ghostworks]

Innovation is extremely overrated. Most of our tech-driven culture is not based on innovation. Not even close. All the innovating was done decades ago, when people started dreaming up what might be possible given phenomena they had only a slippery grasp of once they were leveraged into machines that hadn't yet been built.

No, our culture is driven by cost-reduction. Once something becomes cheap enough, we do it. If it's not, we put it on a shelf until it is. For the computing revolution, the internet bubble of the 90s, and the social web of today, the driving factors have all been about scale: quantity as a quality all its own. You could have done the same things decades earlier -- hell, we _were_ doing the same thing decades earlier -- but only when it's cheap enough that we increase the scale by three orders of magnitude or more did it become a game-changer.

There are plenty of cases in recent history of a product being invented "before its time". It exists in relative obscurity, figuratively collecting dust until it is brushed off to solve the problem that we only got to at just the present moment. You can't drive technology from behind, it has to be pulled along by the context of the situation. Technology by its very nature is very light on innovation, because it is firmly rooted in a practical context of costs and needs.

Now as to whether some technological innovation can be automated: definitely. All engineering is about working around the edges of a problem until you can describe what you need in terms of what you can do. You probe the problem, consider edge cases, and trace the general shape of a missing block until you can say, "what we really need here is some way to measure X and Y, figure out which is closer to Z, and then just give us that one." And then you have a block-level spec. And then you drill into that block as a device all its own, filling in the parts that are obvious from experience and education until you get to the difficult part, the truly novel part, and again say "what we need here is some what to measure X...." Once you get down to a block that only has things that have already been invented, you're done. Then you have your new block, in your new device, solving your new problem. That's what invention is: work.

Get a "smart" enough robot, one which is flexible enough in its model of the world that you can teach it roughly like a human, and you can certainly train it the way you could a junior engineer. A simpler robot will have more limits, but can also be trained to do simpler, smaller steps of the same process. TFA basically describes a system for making a bot context aware... by trawling through records of what humans have done, it can recognize problems that humans have solved before to help another human solve a similar problem (even if he doesn't know it's similar). It can recognize that a margarita machine, a cement mixer, and washing machine all have similar problems to solve on some level, even though no one human really looked at any two of those problems. It looks at your statement of "what we need here...," and chimes in "oh, like a ____ but for _____."

Re:Mathematics

[jewens]

Yes but this is with a computer! (on the internet?)

Should You Use the Betteridge Cliche?

[Stirling Newberry]

Is the correct formulation here.

Analogy Finder?

[ilsaloving]

I tried to use that analogyfinder site, but I couldn't find the setting for 'cars'.